Atomistic simulations of GeSbTe alloys for applications in electronic phase change memories

The scope of this thesis is the study of some properties of GeSbTe alloys by means of atomistic simulations. These alloys are classified as phase change materials which are employed in data storage applications, such as a novel type of electronic non-volatile memory called Phase Change Memory (PCM). The prototypical alloy used so far in PCM is the Ge2Sb2Te5 (GST225) compound. In recent times, PCM has also gained significant attention for embedded memories for automotive applications. However, the crystallization temperature Tx (420 K) of GST225 is too low for applications that require data retention at higher temperatures. Therefore, other options have been investigated for these applications. Ge-rich GeSbTe alloys have emerged as the most promising materials for embedded memories. The Tx of the Ge-rich GST alloy is higher compared to the stoichiometric alloy GST225, because of phase separation with segregation of Ge atoms during crystallization. However, atomistic details of the phase separation and crystallization processes of Ge-rich GeSbTe alloys are not fully understood. In this respect, atomistic simulations could provide details at the atomistic level of both the crystallization and segregation phenomena. On these premises, in this thesis we addressed by atomistic simulations three different issues on the properties of phase change materials of interest for applications in PCM. Firstly, we calculated the electrical and thermal conductivities and the Seebeck coefficient of the liquid phase of GST225 by molecular dynamics simulations based on density functional theory (DFT), since these coefficients are needed to perform an electrothermal modeling of the PCM. Secondly, we studied the effect of H/N doping on the atomic mobility in Ge5Sb2Te3 and Ge2Te by DFT molecular dynamics in the liquid e supercooled liquid phase. Indeed, in literature it was reported that presence of N could further raise the Tx of Ge-rich GeSbTe alloys probably due to the reduction of Ge mobility, which slows down the segregation of Ge during the crystallization process. However, unintentional doping of H during the deposition process could possibly spoil the effect of N atoms. Finally, in the perspective to develop an interatomic potential for Ge-rich GeSbTe alloys to study the crystallization mechanism and phase separation process by large scale molecular dynamics, we have generated a machine learning interatomic potential for the binary Ge-rich GexTe alloys by fitting a database of DFT energies and forces. We decided to study the binary alloy because it shares several properties with the ternary system but it is easier to model. We also used this potential to study the crystallization mechanism and the phase separation process in the Ge2Te alloy by large scale molecular dynamics simulations lasting up to 14 ns.